Joshua D. Drake wrote:
On Wed, 23 Apr 2008 16:27:33 +0100
John Gardner <john.gardner@xxxxxxxxxxxx> wrote:
We have two PostgreSQL servers (8.2) running in a cluster.
Could you be a bit more specific about what you mean by: in a cluster?
Well, we're using middleware technology to load balance and cluster the
PostgreSQL backends
Now, we're not seeing any problems with performance and we're not
seeing any bloat but I have a couple of questions which I hope some
of you can help me out on.
1) Is using the autovacuum daemon and running vacuumdb from a cron
job overkill?
Not necessarily. Autovacuum can get behind when managing larger tables
that are actively used. This is better in 8.3 because you can have more
than one autovacuum worker.
2) The vendor of the clustering software suggests that I regularly
run vacuum --all --analyze on a specific table. Surely, if I'm
running vacuumdb --all --analyze isn't this just a wrapper for
running the vacuum --all --analyze command anyway?
Well you have stated two different things here.
1. The vendor is stating you should vacuum a specific table
2. You are vacuuming the whole database.
Do you have a specific table that they are talking about? Is it
particularly busy? Say a session table?
OK, well initially, the vendor told us to stagger the vacuumdb command
every couple of hours, but latterly they are telling us to vacuum a
specific table. The table contains the transactions that are used for
rollback if any particular node in the cluster needs to be restarted.
I guess that I'm
asking why would I want to run vacuum --all --analyze on a specific
table, when I regularly run vacuumdb --all --analyze on all databases?
Thanks
John
